tech
Introducing Mistral OCR 3
Breakthrough performance: 74% overall win rate over Mistral OCR 2 on forms, scanned documents, complex tables, and handwriting.
TL;DR
- Mistral OCR 3 extracts text and embedded images from various documents with high fidelity.
- It supports markdown output with HTML-based table reconstruction for understanding document structure.
- The model is smaller and more affordable than competitors, priced at $2 per 1,000 pages with a 50% Batch-API discount.
- It excels in processing handwriting, forms, scanned/complex documents, and intricate tables.
- Mistral OCR 3 is a significant upgrade from Mistral OCR 2 across all languages and document types.
- It can be integrated via API (mistral-ocr-2512) or used through the Document AI UI.
- Early customers use it for invoice processing, digitizing archives, and extracting data from reports.
- A self-hosting option is available for organizations with strict data privacy requirements.
- OCR is foundational for generative and agentic AI, providing richer context from data.